Living on the Edge Opportunities for Amur Tiger Recovery in China

Home , Siberian roe deer

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/321183153

Living on the edge: Opportunities for Amur tiger recovery in China

Article in Biological Conservation · January 2018 DOI: 10.1016/j.biocon.2017.11.008

CITATIONS READS 0 200

12 authors, including:

BNU_tianming Wang J. andrew Royle Beijing Normal University United States Geological Survey

38 PUBLICATIONS 131 CITATIONS 280 PUBLICATIONS 13,007 CITATIONS

SEE PROFILE SEE PROFILE

James l.D. Smith Jianping Ge University of Minnesota Twin Cities Beijing Normal University

77 PUBLICATIONS 2,462 CITATIONS 70 PUBLICATIONS 483 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

wildlife protection along roads in Changbai Mountain View project

The development of metapopulation models View project

All content following this page was uploaded by BNU_tianming Wang on 21 November 2017.

The user has requested enhancement of the downloaded file. Biological Conservation 217 (2018) 269–279

Contents lists available at ScienceDirect

Biological Conservation

journal homepage: www.elsevier.com/locate/biocon

Living on the edge: Opportunities for Amur tiger recovery in China T Tianming Wanga,d,e, J. Andrew Royleb, James L.D. Smithc, Liang Zoua,d,e, Xinyue Lüa,d,e, Tong Lia,d,e, Haitao Yanga,d,e, Zhilin Lia,d,e, Rongna Fenga,d,e, Yajing Biana,d,e, Limin Fenga,d,e, ⁎ Jianping Gea,d,e, a Ministry of Education Key Laboratory for Biodiversity Science and Engineering, Beijing Normal University, Beijing 100875, China b U.S. Geological Survey, Patuxent Wildlife Research Center, Laurel, MD 20708, USA c Department of Fisheries, Wildlife and Conservation Biology, University of Minnesota, St Paul, MN 55108, USA d State Key Laboratory of Earth Surface Processes and Resource Ecology, Beijing Normal University, Beijing 100875, China e College of Life Sciences, Beijing Normal University, Beijing 100875, China

ARTICLE INFO ABSTRACT

Keywords: Sporadic sightings of the endangered Amur tiger Panthera tigris altaica along the China-Russia border during the Amur tiger late 1990s sparked efforts to expand this subspecies distribution and abundance by restoring potentially suitable Habitat use habitats in the Changbai Mountains. To guide science-based recovery efforts and provide a baseline for future Density monitoring of this border population, empirical, quantitative information is needed on what resources and Occupancy modeling management practices promote or limit the occurrence of tigers in the region. We established a large-scale field Cattle grazing camera-trapping network to estimate tiger density, survival and recruitment in the Hunchun Nature Reserve and Transboundary conservation the surrounding area using an open population spatially explicit capture-recapture model. We then fitted an occupancy model that accounted for detectability and spatial autocorrelation to assess the relative influence of habitat, major prey, disturbance and management on tiger habitat use patterns. Our results show that the ranges of most tigers abut the border with Russia. Tiger densities ranged between 0.20 and 0.27 individuals/100 km2 over the study area; in the Hunchun Nature Reserve, the tiger density was three times higher than that in the surrounding inland forested area. Tiger occupancy was strongly negatively related to heavy cattle grazing, human settlements and roads and was positively associated with sika deer abundance and vegetation cover. These findings can help to identify the drivers of tiger declines and dispersal limits and refine strategies for tiger conservation in the human-dominated transboundary landscape. Progressively alleviating the impacts of cattle and human disturbances on the forest, and simultaneously addressing the economic needs of local communities, should be key priority actions to increase tiger populations. The long-term goal is to expand tiger distribution by improving habitats for large ungulates.

1. Introduction mostly been subdivided into smaller, less viable populations (Carroll and Miquelle, 2006; Kenney et al., 2014; Rayan and Linkie, 2015; Asian forest ecosystems are becoming increasingly fragmented by Walston et al., 2010). To better understand the consequences of habitat the extensive intensification of anthropogenic activities (Joshi et al., fragmentation and degradation, recent conservation research has fo- 2016; Z.W. Li et al., 2009; Wang et al., 2012). As a result, tigers cused on how tigers use and disperse through human-dominated (Panthera tigris) have declined steadily since the end of WWII and are on landscapes (Carter et al., 2012; Chanchani et al., 2016). the brink of local extinction in many areas (Gopal et al., 2010; Walston Attempts at recovering small, threatened populations of Asian car- et al., 2010). Tigers, a flagship species and the apex predator in Asia, nivores often involve expanding their range beyond extant protected exemplify the problems faced by most large carnivores worldwide; they areas and international borders (Carter et al., 2012; Chanchani et al., have experienced substantial population declines and range contrac- 2016; Linnell et al., 2016; Smith et al., 1998). In Northeast China re- tions during the past century (Dinerstein et al., 2007). Despite decades covery of tigers will require expanding conservation efforts beyond the of conservation actions, human land use has led to a steady loss of border with Russia into landscapes where human activities and biodi- habitat and as a result, once large and continuous populations have versity conservation must be integrated. This is the case with the

⁎ Corresponding author at: No. 19, Xinjiekouwai Street, Haidian District, Beijing 100875, China. E-mail addresses: [email protected] (T. Wang), [email protected] (J. Andrew Royle), [email protected] (J.L.D. Smith), [email protected] (L. Feng), [email protected] (J. Ge). https://doi.org/10.1016/j.biocon.2017.11.008 Received 7 January 2017; Received in revised form 27 October 2017; Accepted 6 November 2017 0006-3207/ © 2017 Elsevier Ltd. All rights reserved. T. Wang et al. Biological Conservation 217 (2018) 269–279

Fig. 1. Monitoring areas of the long-term Tiger-Leopard Observation Network (TLON) in NE China showing camera placement relative to settlements, major roads and nature reserves or national parks. Red dots represent the sample locations (camera traps) where tigers were observed. (For interpretation of the references to color in this ﬁgure legend, the reader is referred to the web version of this article.)

endangered Amur (Siberian) tiger (P. t. altaica), which occurs on the infrastructure development, and forest logging have seriously jeo- northern fringe of the tiger's range. This subspecies plays a vital role in pardized their viability (Tian et al., 2011; Wang et al., 2016). In par- structuring the mixed coniferous and broad-leaved forest ecosystems of ticular, changes in land-use policies in the Changbai Mountains have Northeast Asia (Luo et al., 2004; Miquelle et al., 2010b). Historically, led to an increase in cattle ranching over the past 20 years, which has tigers were once distributed widely across much of Northeast China, the resulted in habitat degradation and the exacerbation of human-tiger Far East of Russia and the Korean Peninsula, but in recent decades they conflicts (Soh et al., 2014; Wang et al., 2016). Thus, a conservation have experienced severe demographic and geographic range contrac- strategy is urgently needed that provides ecological services for human tions due to habitat loss, poaching, prey depletion and disease (Gilbert needs and habitat for tigers and leopards. et al., 2015; Miquelle et al., 2010a; Tian et al., 2011; Wang et al., 2016). Despite recent research that has increased the understanding of Currently < 600 individuals are estimated to remain in two isolated Amur tiger ecological requirements (e.g., habitat connectivity, prey subpopulations confined to the Sikhote-Alin Mountains of Russia (95% availability and human disturbance) (Carroll and Miquelle, 2006; of individuals) and the Changbai Mountains along the China-Russia Hebblewhite et al., 2014; Miquelle et al., 2010b; Petrunenko et al., border (5%) (Miquelle et al., 2006; Tian et al., 2009). Movement be- 2016), significant knowledge gaps remain. In particular, programs that tween the two subpopulations is blocked by an urbanized rail and prioritize the initiation of such recoveries should be evidence-based, highway corridor and wetlands (Carroll and Miquelle, 2006; requiring assessments of landscape-wide conditions for this population, Hebblewhite et al., 2014; Miquelle et al., 2015), resulting in genetic which are typically unclear in China. To improve the conservation divergence (Henry et al., 2009; Sorokin et al., 2016). This subspecies outlook for the tiger, in 2016 the Chinese government initiated a Tiger- has the lowest genetic variation of all extant tiger subspecies and the Leopard National Park (TLNP) program to expand the Amur tiger and Southwest Primorye/Changbai Mountains population has an estimated leopard ranges in China (McLaughlin, 2016). The TLNP is connected to effective population size of 11–14 (Alasaad et al., 2011; Dou et al., the Land of Leopard National Park (LLNP) in southwest Primorye Krai, 2016; Henry et al., 2009). The future of the Amur tiger, especially this Russia, but on the Russian side of the border, habitat is limited. In smaller, isolated transboundary population is at a crucial threshold. contrast, in China, there is extensive potential habitat, but wildlife face Adequate conservation efforts are needed in Northeast China to restore pressure from rapid land-use changes and high levels of anthropogenic landscape permeability so that tigers can once again occupy what is activities. In particular, peripheral anthropogenic encroachment and now a mixture of natural and human dominated habitat (Pitman et al., activities have largely confined tigers to the reserve (Wang et al., 2016). 2017; Yumnam et al., 2014). Hence, understanding how tiger abundance and habitat use in China Since the late 1990s, this border population has gradually increased vary in response to environmental and anthropogenic factors and ex- and is extending its distribution into China (T.M. Wang et al., 2015; isting land management practices (i.e. livestock grazing) is essential for Wang et al., 2016; Wang et al., 2014). This transboundary population re-establishing tigers in China. There is also a clear need to identify how shares land with the only remaining population of the Amur leopard protection designation (inside and outside protected area) influence (Panthera pardus orientalis). These carnivores currently compete with tiger occurrence at local and landscape scale, to inform land-use plan- local people for limited resources (e.g., food) as agricultural expansion, ning and wildlife management.

270 T. Wang et al. Biological Conservation 217 (2018) 269–279

In this study, we used 12 months of camera trapping data across a photographs 24 h/day with a 1-minute interval between consecutive ~5000 km2 landscape to elucidate Amur tiger status and the ecological events. Approximately 70% of the stations had two cameras. The correlates that predict their distribution and abundance. We estimated cameras were operated continuously throughout the year. We visited the spatially explicit density of tigers across a gradient of habitat pro- each camera 5–7 times a year to download photos and check batteries. tection and anthropogenic disturbances using an open population spa- We analyzed tigers, their principal wild prey (sika deer, wild boar tially explicit capture-recapture model. We then used occupancy mod- and roe deer), domestic livestock, and human presence (e.g., rural eling, accounting for detectability and spatial autocorrelation, to assess people using the forest and border patrols on foot and vehicles) as the relative influences of habitat, prey, disturbance, and management “entities” in the camera traps. Each tiger was identified both visually on tiger distribution and abundance. The results of our research will and using the pattern-matching software ExtractCompare (Hiby et al., inform science-based conservation strategies for integrating tiger re- 2009). No single-sided camera trap tiger photo was used unless it covery into a regional landscape-scale plan that includes biodiversity matched a photo from a double-sided station. Sex could usually be and ecological services. determined visually. Tiger cubs (< 1 year old) were removed from the density analyses because they usually remain with their mothers and 2. Materials and methods exhibit low capture rates (Barlow et al., 2009; Karanth and Nichols, 1998). We calculated the detection frequency of each entity at each trap 2.1. Study area station as the number of detections per 100 camera-trap days (Carter et al., 2012; O'Brien et al., 2003). To avoid inflated counts caused by This research was conducted in the northern portion of the repeated detections of the same event, only one record of a species at a Changbai Mountains in Jilin Province, China, adjacent to southwestern trap site per 0.5 h was included in the data analysis (O'Brien et al., Primorsky Krai, Russia, to the east, and North Korea to the southwest 2003). We use R package “overlap” to estimate the overlapping of the (Fig. 1). The approximately 5000-km2 study area forms the core of a activity patterns of tigers and human presence (Ridout and Linkie, potential recovery landscape for tigers and leopards in China 2009). (Hebblewhite et al., 2012; Wang et al., 2016; Wang et al., 2017). Ele- vations range from 5 to 1477 m. The climate is characterized as tem- 2.3. Open population models perate continental monsoon with average annual temperatures ranging from 3.90–5.65 °C and a frost-free period of 110–160 days/year (http:// We estimated population parameters using a Jolly-Seber model www.weather.com.cn). The annual average precipitation is (Gardner et al., 2010; Jolly, 1965; Seber, 1965) and a multi-session 580–618 mm, with the most precipitation occurring in the summer model. An spatially explicit capture-recapture (SECR) model is a hier- from June to August. The majority of forests have been logged, and archical model that explicitly links the spatial locations and movements many low-elevation forests have been converted into secondary de- of individuals (the point process) to the imperfect encounters of in- ciduous forests over the past 5 decades (Z.W. Li et al., 2009). dividuals in a trapping array (the observation process) (Efford, 2004; The prey of tigers in this area includes sika deer (Cervus nippon), Royle et al., 2009). Open SECR models allow the joint modeling of data wild boars (Sus scrofa), Siberian roe deer (Capreolus pygargus), musk over periods of time for which closed models may not be appropriate. deer (Moschus moschiferus), Asiatic black bears (Ursus thibetanus), and The accommodate non-closure by incorporating explicit dynamics al- domestic species (cows and dogs), along with small animals such as lowing for individuals entering and leaving the population. The Jolly- Asian badgers (Meles leucurus) and raccoon dogs (Nyctereutes procyo- Seber SECR model developed by Gardner et al. (2010) is based on a noides)(Kerley et al., 2015; Tian et al., 2011; Xiao et al., 2014). The Bayesian Markov chain Monte Carlo (MCMC) framework. This method average home range of the Amur tiger is 401 and 778 km2 for females used data augmentation for the number of individuals alive (N) during and males, respectively, based on data from the adjacent southwestern each year, per capita recruitment (ρ), and apparent survival rate (φ), Primorsky Krai (Hernandez-Blanco et al., 2015). Together, the tiger and where we set augmentation value for N at 150 individuals (Gardner leopard density is < 1 individuals/100 km2 in this area (Wang et al., et al., 2010; Royle and Dorazio, 2012; Royle et al., 2009). This study 2017; Xiao et al., 2016). was conducted during 2 consecutive periods made up of 6 months each Over the past decade, the study area has been exposed to increasing from August 2013 to July 2014. We calculated ρ as the number of new levels of agricultural and industrial development, particularly mining individuals in the year 2014 divided by the number of animals alive in and new road building, which has led to habitat fragmentation. Since the year 2013. We note that these parameters should be interpreted as 2015, commercial logging of natural forests has been halted and forest apparent survival and recruitment because the population is susceptible cover has expanded as part of a national plan to promote ecological to permanent emigration and immigration which affect estimators of development. The main economic activity in rural areas is free-range survival and recruitment, respectively. cattle grazing; other human activities include the collection of edible For the present study, our 6-month sample periods are greater than ferns, ginseng farms, and frog farming (Wang et al., 2016). previously estimated periods of 45 to 90 days necessary for closure for studies of other tiger subspecies (Duangchantrasiri et al., 2016; Karanth 2.2. Data collection and field methods and Nichols, 1998). Based on the much lower density (< 0.5 individuals/100 km2) and lower detection probability of the Amur tigers This study, conducted from August 2013 to July 2014, was part of a compared to other subspecies (Xiao et al., 2016), this longer sampling long-term Tiger Leopard Observation Network (TLON) project that period will be necessary to ensure most individuals are detected two or employed 356 camera trap stations in the Hunchun Nature Reserve more times. Moreover, tigers are not known to shift their home range (HNR) and two newly established reserves, Laoyeling and Wangqing, in seasonally (Smith et al., 1987). A closure test also was conducted within 2014 (Wang et al., 2016)(Fig. 1). We used 3.6 × 3.6 km grids to guide the secr package for each trapping period. Camera trapping data within camera trap placement throughout the study area. On average, there each period was subdivided into 2-week intervals (13 sampling occa- were approximately 20 cameras per female tiger home range. Within sions each). The spatial detection history was constructed according to the sampling grids, we maximized the detection probability by placing whether an animal was photographed during an occasion (Table 1). cameras at sites where tigers, leopards, and their prey are likely to Although monitoring data from neighboring Russia were unavailable, travel (e.g., along ridges, valley bottoms, trails, forest roads and near SECR models can account for the detection of “foreign residents” in scent marked trees). We excluded grids on farmland and in villages. The abundance and density estimates (Bischof et al., 2016). The Jolly-Seber cameras (LTL 6210M, Shenzhen, China) were fastened to trees ap- SECR models were fitted using the jagsUI package (Kellner, 2016) in the proximately 40–80 cm above the ground and were programmed to take R software environment (version 3.3.1, R Development Core Team,

271 T. Wang et al. Biological Conservation 217 (2018) 269–279

Table 1 Summary of Amur tigers captured by camera traps in 2013–2014, showing ID number, gender, the number of capture sites, capture frequencies, maximum distance moved (MDM) and capture history over 26 sampling occasions. Fourteen captures were of insuﬃcient quality to allow individual recognition (not shown).

ID Gender Sites Capture frequencies MDM Sampling occasion

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26

TIG-02 Male 27 79 51.83 1 1 1 1 1 1 1 1 1 1 0111111111101101 TIG-03 Female 8 30 7.51 0 1 1 1 1 1 1 1 0 0 1000100111001011 TIG-04 Female 15 42 26.20 0 0 1 1 1 1 0 1 1 0 1100110001110111 TIG-05 Female 14 14 42.42 0 0 1 0 0 0 1 1 0 0 0001011110010100 TIG-08 Male 15 56 31.72 1 1 1 1 1 1 1 1 1 1 0111011111011011 TIG-09 Female 5 15 21.46 0 0 0 0 1 1 1 1 0 0 1001010010010100 TIG-10 Male 1 1 – 00100000000000000000000000 TIG-11 Male 8 17 23.65 1 1 1 1 1 1 1 1 0 0 0000000000100000 TIG-12 Male 5 9 15.38 0 0 0 1 1 1 1 1 0 0 0000000000000000 TIG-13 Female 6 13 9.49 1 0 1 1 1 1 1 0 0 0 1000000000000011 TIG-14 Female 1 1 – 00000000000000000000000100 TIG-15 Male 3 3 7.56 0 0 0 0 0 0 0 0 0 0 0000000000010000 TIG-17 Male 3 3 10.72 0 0 0 0 0 0 0 0 0 0 0000000000000010 TIG-20 Male 11 24 42.19 0 0 0 0 0 0 0 0 0 0 0001111111111110 TIG-22 Female 9 29 15.98 0 0 1 1 1 1 0 1 1 1 0100101110001111 TIG-23 Male 3 3 14.15 0 0 0 0 0 0 0 0 0 0 0000000000000011 TIG-24 Female 3 3 14.64 0 0 0 0 0 0 0 0 0 0 0000000000000011

2016). We ran 3 Markov chains with 20,000 iterations each with the use detection-nondetection camera-trapping data from repeat surveys first 5000 discarded as burn-in, for a total of 45,000 posterior samples. to estimate the probability of an animal occurring at least once (Ψ) and We checked the convergence of the MCMC simulation by each para- being detected at a camera site (p). These models account for imperfect meter's trace plot and the Gelman-Rubin statistic (Brooks and Gelman, detection and allow both parameters to vary in response to covariates. 1998). We assessed the model goodness-of-fit using the Bayesian P- Given there were multiple camera trap sites within each tiger's home value, with 0.05 < P < 0.95 indicating model adequacy (Gelman range, it is habitat use, rather than occupancy that we are modeling et al., 1996). (Linkie et al., 2006; Mackenzie and Royle, 2005). By sampling year To compare the tiger density inside and outside the HNR (i.e., core round, we also simplified the interpretation of occupancy probability as population near the border vs. the more dispersed population) in 2013 the proportion of area used by tigers. We assumed animals move ran- and 2014, we also ran a maximum likelihood-based multi-session SECR domly between the fine-scale sampling sites, which relaxed the as- model using the R package secr (Efford, 2016). We specified years as sumption of geographical closure typically required for occupancy different sessions and sex as a group factor. This allowed fitting the models. In our study, we suspected autocorrelation between adjacent model with detection parameters (baseline encounter rate, λ0, and scale camera locations, which were an average of 2.36 km apart. Thus, to parameter, σ) shared across year and sex to obtain a more precise es- avoid biased estimates of model parameters, we used a hierarchical timate for the year that had the lowest number of captures. The camera spatial occupancy model, which explicitly incorporates a spatial traps were treated as a count detector type that allowed for repeat random effect and employs a probit link function to increase compu- detections of the same individual at the same camera station per oc- tational efficiency (Johnson et al., 2013). We first identified supported casion (Efford et al., 2009). Density models were fitted in secr using full maximum likelihood-based occupancy models with the unmarked likelihood with a hazard half-normal function. package in R (Fiske and Chandler, 2011) and then fit Bayesian versions We constrained the state-space based on known tiger occupied of the selected models using the stocc package in R (Johnson et al., ranges (Wang et al., 2016). The sea is approximately 20 km to the east 2013). and in the west, few tigers were observed from our camera-trap survey We defined 2-week periods as temporal replicates and constructed area. For both methods, we therefore used a 20-km buffer width around tiger detection histories for each camera site over 23 sampling occa- the camera trapping grid and established 2 km × 2 km cells using sions. Next, we explored a set of biotic and abiotic covariates that re- ArcGIS 10.1 (ESRI, Redlands, CA, USA) as the state-space. We also pre- presented hypothesized ecological relationships of habitat for tigers evaluated a larger buffer (40-km) but found this 20-km buffer should be (e.g., habitat structure, prey availability, disturbance and management) sufficient to contain all active centers of detectable individuals. Even if (Table 2)(Carter et al., 2012; Chanchani et al., 2016; Harihar et al., this buffer size may not be large enough to account for transients, a 2014; Hebblewhite et al., 2014; Miquelle et al., 2015; Petrunenko et al., recent study shows that density estimates of SECR models are robust 2016). Here, given rapid forest policy changes could greatly reduce even when a fairly large number of transients occur in the population cattle abundance in short term, we defined cattle as a management during the sampling period (Royle et al., 2016). We excluded any non- factor. Initially, we considered 10 variables as predictors of tiger oc- forest habitat within the state-space and assumed that the observed cupancy and 5 as predictors of detection. Because tiger occurrence is individual- and camera-specific encounter frequencies follow the influenced by topography, we derived 3 topographic covariates from Poisson encounter model with no temporal or individual behavior the Shuttle Radar Topography Mission (SRTM) 30 m digital elevation variation in detection probability (see Royle et al., 2009 for details). To model. These were elevation, vector roughness measure (VRM; e.g., improve the estimates of detection probability, we accounted for local variation in elevation change) (Sappington et al., 2007) and to- varying effort when cameras were not functioning from errors, damage pographic position index (TPI; e.g., finer scale depressions or ridges) from cattle, or interference by humans. (De Reu et al., 2013). The VRM and TPI were calculated using a circular neighborhood with a 1-km radius. The normalized difference vegetation index (NDVI), derived from the 250 m Moderate Resolution Ima- 2.4. Occupancy models ging Spectroradiometer (MODIS) imagery (product MOD13Q1) of the study area, was used as a proxy for vegetation productivity and cov- We assessed the habitat use of tigers across the study area using erage (Harihar et al., 2014; Pettorelli et al., 2016; Wang et al., 2012). single-season occupancy models (MacKenzie et al., 2002). These models

272 T. Wang et al. Biological Conservation 217 (2018) 269–279

Table 2 Variables used for occupancy models to model habitat use by Amur tigers.

Name Description Categories Source Parameter and expected inﬂuence

Elevation (Elev) Numeric (m), elevation of point generated Habitat Shuttle Radar Topography Mission (SRTM) Ψ (−) from 30 m DEM 1 Arc-Second Globala TPI Numeric, topographic position index Habitat Measured from elevation gridsb Ψ (−) VRM Numeric, vector ruggedness measure Habitat Measured from elevation grids Ψ (+) NDVI Numeric, normalized diﬀerence vegetation Habitat MODIS Vegetation Indices 250 m 16 day Ψ (+) index NDVI (MOD13Q1)c Distance to border (Dist.border) Numeric (m), distance to the nearest border Habitat China Fundamental Geographic p (−) from camera Information Dataset Trail Categorical, valley forest road or ridge trail Habitat Field sampling p (+) Wild boar Numeric, detection frequencies of each prey Prey Camera trap Ψ (+) Roe deer species per 100 trap-days Sika deer Human presence (Human) Numeric, detection frequencies of human on Disturbance Camera trap p (−) foot traﬃc per 100 trap-days Distance to road (Dist.road) Numeric (m), distance to the nearest road from Disturbance Local Forest Resource Distribution Map Ψ (+), p (+) camera Distance to settlement Numeric (m), distance to the nearest settlement Disturbance China Fundamental Geographic Ψ (+), p (+) (Dist.settlement) from camera Information Dataset Cattle Numeric, detection frequencies of cattle per Management Camera trap Ψ (−) 100 trap-days

a SRTM dataset (https://lta.cr.usgs.gov/SRTM1Arc). b Topography Tools for ArcGIS 10.1 (http://www.arcgis.com/home/item.html?id=b13b3b40fa3c43d4a23a1a09c5fe96b9). c MODIS vegetation indices (https://lpdaac.usgs.gov/dataset_discovery/modis/modis_products_table/mod13q1).

We calculated the camera site-level average NDVI during 2010–2014 regression (RSR) using the stocc package (Hughes and Haran, 2013; for modeling tiger occupancy. We used the detection frequencies from Johnson et al., 2013) and considered all combinations of informative our camera traps as surrogates for the abundance of each prey species covariates to create a full multivariable model that included the com- (wild boar, roe deer and sika deer). The detection frequencies for bined effects. Subsequently, we assessed the relative importance of groups of cattle and humans were also taken from the camera detection habitat and prey vs. disturbance and management on tiger occurrence. data as a quantitative measure of human use and grazing intensity. We For these models, only covariates that appeared in the highly supported also used ArcToolbox in ArcGIS 10.1 to calculate the distances to models from the first stage were used. We expected higher support for landscape features that may affect the probability of tiger occurrence the model that combined habitat and prey models compared to the (e.g., distance from each camera to a settlement, road or international model that combined disturbance and management. We set the distance border). During camera deployment, we recorded the trail type (forest threshold for detecting spatial structure in neighboring sample loca- dirt road or ridge) at each camera location. All continuous covariates tions at 12 km based on the average size of female tiger home ranges were transformed into standardized z-scores to facilitate the inter- and the spatial distribution of camera traps (Hernandez-Blanco et al., pretation of the covariate coefficients and to improve model con- 2015). We specified flat prior distributions for both the detection and vergence. A variance inflation factor (VIF), which measures multi- occupancy processes and a gamma (0.5, 0.0005) distribution for the collinearity among variables, was calculated for all of the covariates, spatial component following Johnson et al. (2013). The Moran cut used and covariates with a VIF < 3 were retained in the model. Pearson's in the spatial model was 10% of the number of sites. We allowed the correlation coefficients were also calculated to further check for evi- chain to stabilize by running the Gibbs sampler for 400,000 iterations dence of collinearity, and when the correlated variables were with a burn-in of 100,000 iterations. Every 50th sample was retained |r| > 0.7, one variable was excluded from the same model. for a total of 6000 posterior samples to estimate the parameter mean, We selected models in a 2-step process under a maximum likelihood SD, and 95% Bayesian credible interval (CI). Covariates are considered framework. First, detection probability was modeled as a function of all to have a significant association with tiger habitat use if their 95% CI do combinations of the 5 possible detection covariates while maintaining a not overlap 0. We used the Geweke diagnostic statistics (Geweke, 1992) null reference occupancy model. Subsequently, a candidate set of oc- and the |Z| < 1.96 scores to determine model parameter convergence. cupancy models was created based on the statistical significance of We used the posterior predictive loss criterion (PPLC) (Gelfand and individual occupancy covariates (no interactions) and Akaike's in- Ghosh, 1998) to compare the Bayesian RSR and non-spatial models. We formation criterion (AIC) using a stepwise covariate selection procedure then assessed the robustness of the models using the area under the in the unmarked package followed by the model-fitting process de- curve (AUC) of the receiver-operating characteristic. Similar to Broms scribed in Schuster and Arcese (2013). In this step, we only included the et al. (2014), we calculated the AUC statistic in the ROCR package in R best-performing detection model as derived from the first step. We then using the median occurrences to compare the predicted vs. observed ranked all candidate models by AIC value and considered competitive apparent occupancy among the camera sites. models as those within 2 ΔAIC of the top performing model (Arnold, 2010; Burnham and Anderson, 2012). We then selected the model with 3. Results the fewest parameters within 2.0 ΔAIC of the top model (Arnold, 2010). Because the penalty for adding one parameter is +2 AIC, if only one 3.1. Abundance and density estimates parameter is added but the AIC is within 2 ΔAIC, the model fit was not improved enough to overcome the penalty. We plotted the predicted From August 2013 to July 2014, a total of 356 detections of tigers habitat use probability against each meaningful covariate from our top were obtained over 114,854 trap-days. A total of 21 individual tigers (9 candidate model (ΔAIC < 2) across the range of data while holding all males, 8 females and 4 cubs) were identified from 342 detections by the other covariates at their mean value. their unique stripe patterns. Nine of the adult tigers were present in Next, we modeled spatial autocorrelation with restricted spatial both 2013 and 2014 (Table 1). Each individual was detected an average

273 T. Wang et al. Biological Conservation 217 (2018) 269–279

Table 3 2013 2014 Posterior estimates of open spatially explicit capture-recapture (SECR) model parameters 0.4 for the Amur tiger trapping data. ) Parameter Mean SD 2.50% 50% 97.50% 2 0.3 N1 22.27 4.60 15.00 22.00 33.00

N2 29.14 5.00 21.00 29.00 40.00

D1 0.20 0.04 0.14 0.20 0.30 D2 0.27 0.05 0.19 0.27 0.37 0.2 σ 9.91 0.09 9.67 9.93 10.00

λ0 0.15 0.02 0.11 0.15 0.20 ρ 0.45 0.77 0.27 0.45 0.55 φ 0.83 0.11 0.57 0.84 0.98 0.1 Density (individuals 100 km

Notes: The number of unique individuals observed over both years was 17. N1 and N2 are the number of estimated activity centers in the state-space for 2013 and 2014, respec- 2 tively. Density (D1 and D2) is calculated as animals per 100 km . The per capita re- Inside HNR Outside HNR Inside HNR Outside HNR cruitment is ρ. σ is the scale parameter for the detectability function, which is given in Survey kilometers. λ0 is the basal encounter rate of a tiger whose activity center is located Fig. 2. Tiger density estimates (as the number of individuals per 100 km2) and 95% precisely at a given trap. The apparent survival between the two sampling seasons is φ. confidence intervals calculated from likelihood-based multi-session spatial capture-re- The results are based on 3 Markov chains run for 20,000 iterations each and discarding capture models. Estimates of tiger density inside and outside of the Hunchun Nature the first 5000 as burn-in, for a total of 45,000 iterations. Reserve (HNR) are not independent from one another because tigers are present in both regions. of 20.12 times and at 8.06 different locations across an area of approximately 5000 km2. However, the detection frequencies were highly heterogeneous among individuals and sexes: 3 males (TIG-02, TIG-08 uninformative terms (TPI and VRM) (Table S3). Thus, the following and TIG-20) and 3 females (TIG-3, TIG-4 and TIG-22) accounted for covariates: elev, NDVI, dist.road, dist.settlement, wild boar, roe deer, 73% of all detections, whereas 2 (TIG-10 and TIG-14) were detected sika deer and cattle were meaningful predictors of tiger occupancy and only once in both years (Table 1). The average maximum distance were used to conduct the Bayesian analysis. fi moved (MMDM, ± SE) was 1.25 times larger for males than for females The RSR model provided a better t than the nonspatial model ff (24.65 ± 5.64 km vs. 19.67 ± 4.51 km). (PPLC: 403.32 vs. 410.08), indicating that the random spatial e ect was The Gelman-Rubin convergence diagnostic of R ≤ 1.1 for each warranted. The AUC value for the full RSR model was 0.89, indicating parameter indicated that the Jolly-Seber SECR convergence was ade- that we correctly predicted tiger's occupancy. Tigers noticeably pre- quate, and the Bayesian P-value of 0.62 indicated that the model ade- ferred sites at a lower elevation (with use rapidly declining beyond quately described the data. The posterior mean of the apparent survival 800 m) and with a higher NDVI (Table 4 and Fig. 3). Furthermore, tiger rate (φ) was 0.83, and the posterior mean for the apparent per capita occurrence exhibited a positive association with sika deer abundance. recruitment (ρ) was 0.45 (Table 3). The estimated population sizes (i.e., Finally, tiger habitat use strongly increased as cameras were set farther fi the number of activity centers) for the state-space were 22 (95% away from settlements and roads and was signi cantly negatively re- CI = 15–33) and 29 (95% CI = 21–40) for 2013 and 2014, respectively lated to heavy cattle grazing (Table 4 and Fig. 3). Contrary to ex- (Table 3). Tiger density was 0.20 adult individuals/100 km2 (95% pectation, the model with disturbance and management-related cov- CI = 0.14–0.30) in 2013 and 0.27 individuals/100 km2 in 2014 (95% ariates better explained the probabilities of tiger habitat use than the CI = 0.19–0.37) (Table 3). The closure test supported the assumption model with habitat and prey covariates (Table 5). A map based on the of population closure during the second 6-month sampling period in RSR model shows that sites with a higher probability of habitat use 2014 (z = −0.96, P = 0.17) but did not support the assumption in were concentrated along the border and in the southwest parts of the fi 2013 (z = −2.75, P = 0.003). The posterior mean of the baseline en- study area, generally tting well with tiger observations across space counter rate, λ0, i.e., the expected capture frequency of an individual whose activity center is located precisely at a trap location, was 0.15. The estimated posterior mean for the movement parameter σ was Table 4 The parameter estimates and 95% credible intervals (CI) from a spatial occupancy model 9.91 km (Table 3). for the Amur tiger in NE China. Estimates of beta coefficients are reported for standar- The multi-session SECR model showed that the border core popu- dized covariates, scaled to mean and standard deviation. Covariates with their 95% CI not lation density (0.20 ± 0.06, mean ± SE) inside the HNR was 3.3 encompassing zero (marked in bold) were considered to have a significant association times higher than that outside the HNR (0.06 ± 0.02) in 2013 and was with tiger habitat use and detection. We used the Geweke diagnostic statistics and the |Z| < 1.96 score to determine model convergence. See Table 2 for variable definitions 3.6 times higher in 2014 (0.25 ± 0.07 vs. 0.07 ± 0.02) (Fig. 2). The and abbreviations. male baseline encounter rate, λ0, inside the HNR (0.20 ± 0.03) was 6.7 times higher than that outside the HNR (0.03 ± 0.01), and the Model component Covariate Mean SD 95% CI Z score female baseline encounter rate was 5 times higher (0.10 ± 0.02 vs. Habitat use (Intercept) −0.87 0.56 (−2.27, −0.01) 0.43 0.02 ± 0.006) (Table S1). Elev −1.97 0.93 (−4.55, −0.94) −0.45 NDVI 0.89 0.56 (0.21, 2.42) 0.45 Dist.settlement 1.02 0.65 (0.25, 2.78) 0.47 3.2. Determinants of tiger occupancy Dist.road 1.20 0.75 (0.31, 3.24) 0.16 Wild boar −0.17 0.29 (−0.85, 0.34) 0.59 All covariates were retained because no significant collinearity was Roe deer −0.56 0.49 (−1.77, 0.16) −1.19 Sika deer 1.08 0.71 (0.04, 2.86) −0.48 detected (VIF < 3 and r < 0.7) (Fig. S1). Under a maximum like- Cattle −1.09 0.72 (−2.95, −0.18) −0.92 Ψ lihood framework, the top-ranked detection model, (.) p (trail Detection (Intercept) −2.19 0.15 (−2.48, −1.9) −1.49 + dist.road + dist.settlement + dist.border + human), was used in Dist.border −0.36 0.10 (−0.56, −0.18) −1.27 the subsequent habitat use analyses (Table S2). Six models of tiger Dist.settlement 0.25 0.05 (0.15, 0.35) 1.30 − − − habitat use were competitive (i.e., ΔAIC < 2); following Arnold Dist.road 0.29 0.06 ( 0.40, 0.18) 0.64 ff Trail 0.49 0.14 (0.20, 0.77) 0.07 (2010), we considered models 4 to 6, which di ered by one additional Human 0.07 0.03 (0.02, 0.12) 0.14 parameter but were within two AIC of the top model, to contain 2

274 T. Wang et al. Biological Conservation 217 (2018) 269–279

Fig. 3. Tiger occupancy probability with respect to eleva- 1.0 1.0 tion, NDVI, distance to road, distance to settlement, abundance of sika deer, and abundance of cattle (detec- 0.8 0.8 tions per 100 trap-days). The intervals represent 95% credible intervals. 0.6 0.6

0.4 0.4

0.2 0.2

0.0 0.0 Occurrence probability ( ) Occurrence probability ( ) 317 532 748 964 1179 0.68 0.7 0.72 0.74 0.76 0.78 0.8 Elevation (m) NDVI

1.0 1.0

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0.0 0.0 Occurrence probability ( ) Occurrence probability ( ) 02357 3691215 Distance to road (km) Distance to settlement (km)

1.0 1.0

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0.0 0.0 Occurrence probability ( ) Occurrence probability ( ) 1 4 7 10131619 4 2341607897 Abundance of sika deer Abundance of cattle

Table 5 4. Discussion A summary of the restricted spatial use model for the Amur tiger, with p (trail + dist.- road + dist.settlement + dist.border + human), in NE China. 4.1. Tiger population abundance and density

Model Occupancy covariates PPLC AUC We present the first study of the ecological and anthropogenic Disturbance and dist.road (+), dist.settlement (+), 402.38 0.89 correlates that influence the distribution and abundance of a small management cattle (−) population of Amur tigers. Twelve months of camera trap data were Habitat and prey elev (−), NDVI (+), sika deer (+) 406.34 0.87 analyzed using an open population model. Our study differed from

Notes: Symbols: + and −, covariate exerting a positive or negative effect on tiger oc- other camera trapping studies in that cameras were employed con- currence, respectively. PPLC is the fitted model's minimum posterior predictive loss tinuously for 12 months. We used this approach for two reasons. First, it (PPLC), and AUC is the area under the curve of the receiver-operating characteristic. allowed us to use an open model to account for fluctuations stemming from demographic changes related to adult survival and subadult dis- (Fig. 4). persal (Duangchantrasiri et al., 2016; Goodrich et al., 2008). And The tiger detection probabilities decreased as cameras were placed second, continuous sampling is also part of a long-term monitoring farther from the border (Table 4 and Fig. S2). Tigers were more likely to strategy to track the response of tigers to management actions seeking be detected at locations closer to forest roads, away from settlements, to improved tiger habitat. Because tigers demonstrate strong multi- and with higher abundances of human presence (i.e., high levels of seasonal fidelity to a single home range (Hojnowski et al., 2012; human presence did not decrease p)(Table 4 and Fig. S2). However, Miquelle et al., 1999; Smith et al., 1987), the open model provides tigers were less active during the day when human activity peaked (Fig. information on apparent population survival and recruitment. Our S3). The estimated temporal overlap coefficient between tiger and Jolly-Seber SECR model showed a per capita recruitment of 0.45, which human presence was only 0.31 (95% CI: 0.26–0.35%). With all cov- is consistent with the habitat continuity between our study area and the ariates set to their mean, the tiger detection probability across the study adjacent LLNP in Russia. Given lack of population closure during the area was 0.12 (SE = 0.01). Tigers were estimated to occur across 35% first 6-month trapping period, this violation likely, in part, reflects (95% CI: 28–42%) of the camera trap sites, which is 20% greater than movement of tigers living along the border and thus could lead to the the naïve estimate of habitat use. likelihood of biased estimates of density in 2013. Future studies could sample more sites from both countries to reduce the length of sample periods and thereby increase precision and confidence in inferences. Eight adults were continuously detected for only 1–12 of the

275 T. Wang et al. Biological Conservation 217 (2018) 269–279

Fig. 4. Predicted habitat use probabilities and standard error (SE) for Amur tigers along the China-Russia border estimated by the restricted spatial regression (RSR) model.

sampling occasions (< half a year) (Table 1). Nine individuals were number of livestock killed but identified few tiger locations in areas observed throughout the whole study period; 6 were adult females. If a with high livestock density. These findings suggest that these areas of female remains in the same area for two years, this is strong evidence high livestock density may function as attractive sinks or ecological that she is territorial and therefore a resident breeding animal (Smith traps (Kanagaraj et al., 2011). Similar observations were reported in et al., 1987). Based on 2-year residency, five of 8 females along the Rajaji National Park, India, where tigers occurred at low densities under border are resident breeding animals. The sixth resident female (TIG- anthropogenic influences (intensive forest resource extraction and 05) was photographed near the border in 2012 with sub-adult young, grazing) (Harihar et al., 2009; Harihar et al., 2011). and she was subsequently photographed in 2013 with a new litter of 4 cubs approximately 20 km inland from the border. Since then, she has 4.2. Determinants of tiger habitat use ranged from 10 to 50 km from the border. Her 3-year-old daughter (TIG-22) from her previous litter has taken over her former territory Not surprisingly, tiger habitat use is influenced by the availability of along the border (Wang et al., 2016; Wang et al., 2014). Strong evi- prey, topography and vegetation characteristics. The importance of dence that 6 females reside in our study area also includes photographic elevation on tiger occupancy could be driven by the fact that wild evidence of frequent territorial behavior (e.g. scraping, urine spraying, ungulate prey occur mostly in the low elevation forest. Tigers also claw marking); this documentation highlights the role of the region as a preferred areas with higher NDVI. These results agree with potential stronghold for the recovery of Amur tigers in China. Hebblewhite et al. (2014) and Rayan and Linkie (2015), who found that 2 Tiger densities ranged between 0.20 and 0.27 individuals/100 km tigers preferred lower elevation forest and higher NDVI in Russia and over the study area (Table 3), which is considerably lower than den- Malaysia, respectively, as such areas are most likely favored by their – sities reported for tigers in South and southern Asia (1 18 individuals/ prey. In Russia, resource selection modeling has revealed that the oc- 2 100 km )(Barlow et al., 2009; Carter et al., 2012; Duangchantrasiri currence of principal prey species increases at lower elevation and that et al., 2016; Karanth et al., 2004). Our density estimates are also gen- prey select oak and Korean pine mixed forests with high NDVI values, erally lower than Russian estimates in Ussuriiskii Nature Reserve reflecting their dependence on nuts or acorns and higher canopy cover, – 2 (0.11 0.59 individuals/100 km )(Hernandez-Blanco et al., 2013)orin which occur at lower elevations (Carroll and Miquelle, 2006; – 2 Sikhote-Alin Biosphere Zapovednik (0.15 0.93 individuals/100 km ) Hebblewhite et al., 2014). Thus, it is critical that conservation actions (Soutyrina et al., 2013). target lower elevation forest (< 800 m). Most tigers were photographed inside the HNR within 5 km of the Our results suggest that human activities and increased occurrence southeast border (Fig. 1). We recorded the lowest density estimate of livestock reduces the probability of tiger occupancy outside of HNR. 2 (< 0.1 animals/100 km ) outside the HNR (Fig. 2). In the north and Tigers select habitats farther away from roads and human settlements, southwest of the study area, heavy cattle grazing and human activities which confirms our expectations based on previous findings (Carroll severely degraded the forest understories and posed a threat to tigers, and Miquelle, 2006). Many studies emphasize the negative effects of leopards and their principle prey (see discussion below) (Wang et al., human disturbance (e.g., human settlements, roads and livestock) on 2016; Wang et al., 2017). Wang et al. (2016) documented a high big cats through prey depletion, direct poaching, habitat encroachment

276 T. Wang et al. Biological Conservation 217 (2018) 269–279 or decreased connectivity at large spatial scales (Barber-Meyer et al., This research demonstrates that tigers are expanding their range 2013; Bhattarai and Kindlmann, 2013; Joshi et al., 2013; Linkie et al., into China. A recent study (Dou et al., 2016; T.M. Wang et al., 2015) 2006). Roads in the Russian Far East, which provide access between and our long-term camera trapping survey revealed that at least three villages and towns, are reported to reduce Amur tiger survival rates young male tigers have traveled through a series of large forest patches because of collisions with vehicles and increased poaching of both from the border to > 200 km into China. However, reduction in cattle predators and their prey (Goodrich et al., 2008; Kerley et al., 2002). grazing is needed to decrease both human-tiger conflict and competi- Hebblewhite et al. (2014) observed that wild boar, sika deer and roe tion between livestock and sika deer. To better understand the char- deer also avoid areas with high road densities and towns. In our study, acteristics of a tiger-permeable landscape, studies on tiger movement tigers spatially overlapped with people on foot and vehicles at a fine are needed. This information will help managers to secure corridors spatial scale (i.e., higher detection probability on forest roads), perhaps between the current core habitat on the border and forests further in- by using the twilight and night to avoid human disturbance. Increased side China. detection of tigers along forest roads (e.g., abandoned logging roads), Finally, we stress that effective protection of tigers and their habitat however, could also greatly increase access for poachers and, therefore, can only be achieved by state level commitment, legislation, and road closure or access control is needed. management. In 2016, the Chinese government developed a plan for Occupancy modeling indicated that tiger habitat use has a sig- expanding the TLNP landscape to 15,000 km2, which, when combined nificantly negative correlation with the abundance of domestic cattle with the 4000 km2 Land of Leopard Reserve in Russia, provides a land (Table 4). Similar results were also observed in the Central Terai base for securing a viable tiger population (Hebblewhite et al., 2012). Landscape of India, where tiger habitat use declines as human and li- Furthermore, this plan reduces livestock grazing and timber extraction vestock use increases (Chanchani et al., 2016). Currently > 30% of the to increase forest cover and habitat connectivity so that tigers can re- study area is grazed by domestic livestock at an average stocking rate of colonize core forest areas across this landscape. Its scope is similar in 8–12 cattle/km2, and if this expanding economy continues to exert spatial scale to the Terai Arc Landscape in India and Nepal (Chanchani unsustainable pressures year-round on the habitat, wild prey density, et al., 2016; Harihar et al., 2011), and these two landscape approaches especially that of sika deer, will be depressed. Wang et al. (2017) re- have a broad goal of integrating biodiversity objectives and increasing ported that the detection frequency for cattle across our study area was the ecological services that improve human well-being. Our findings 4.0 times greater than that for sika deer. In our study area, cattle, which were provided to the relevant stakeholders to facilitate this process. weigh 400–600 kg, can reduce the plant biomass in the shrub-herb layer by 29–70% (unpublished data). As a consequence, areas in- Acknowledgments tensively grazed by cattle may lower the habitat quality for sika deer, which is the tiger's most common prey in our study area (comprising We sincerely thank Vitkalova Anna Vladimirovna, Somphot 25–54% of the total biomass consumed) (Kerley et al., 2015). This Duangchantrasiri and Saksit Simcharoen for their great help verifying matches earlier findings elsewhere that demonstrated wild herbivores the identities of individual tigers. We also thank the State Forestry are particularly susceptible to competition with livestock by resource Administration, the Jilin Province Forestry Bureau, and the Forestry limitation and even spatial exclusion (Mishra et al., 2002). F. Wang Industry Bureau of Heilongjiang Province for kindly providing research et al. (2015) reported that habitat overlap with cattle within bamboo permits and facilitating fieldwork. We thank Kelly Perkins, Pu Mou and forests limited the distribution of giant pandas (Ailuropoda melanoleuca) 2 anonymous referees for their thoughtful reviews of the manuscript. in China; in India, domestic livestock resulted in a decline in chital (Axis This work was supported by grants from the National Natural Science axis), sambar (Rusa unicolor) and gaur (Bos gaurus), thereby decreasing Foundation of China (31470566, 31210103911, 31421063 and the density of tigers (Dave and Jhala, 2011; Madhusudan, 2004). 31270567), the National Scientific and Technical Foundation Project of Cattle are left unattended to roam freely from spring to fall in our China (2012FY112000), and the National Key Research and study area and thus are also potential prey for tigers. We recorded > 50 Development Program (2016YFC0500106). livestock killed by tigers every year. However, we suspect that humans disturb tigers feeding on their livestock because tigers do not feed on Appendix A. Supplementary data 25% of the livestock they kill and in 42% of the kills less than half the carcass was consumed (B. Li et al., 2009). The current practice of Supplementary data to this article can be found online at https:// paying compensation for livestock depredation by tigers exacerbates doi.org/10.1016/j.biocon.2017.11.008. these conflicts because it reduces the financial risk to herders and en- courages grazing of livestock in the forest, which in turn degrades the References habitat for wild ungulates (Pettigrew et al., 2012; Soh et al., 2014). Despite the negative impact of cattle grazing, we believe that com- Alasaad, S., Soriguer, R.C., Chelomina, G., Sushitsky, Y.P., Fickel, J., 2011. Siberian tiger's pensation should continue in the short term because it reduces re- recent population bottleneck in the Russian Far East revealed by microsatellite markers. Mamm. Biol. 76, 722–726. taliatory killings of tigers and, if livestock are removed, would not be Arnold, T.W., 2010. Uninformative parameters and model selection using Akaike's needed in the long term. Information Criterion. J. Wildl. Manag. 74, 1175–1178. Barber-Meyer, S.M., Jnawali, S.R., Karki, J.B., Khanal, P., Lohani, S., Long, B., MacKenzie, D.I., Pandav, B., Pradhan, N.M.B., Shrestha, R., Subedi, N., Thapa, G., Thapa, K., 4.3. Conservation implications and recommendations Wikramanayake, E., 2013. Influence of prey depletion and human disturbance on tiger occupancy in Nepal. J. Zool. 289, 10–18. In conclusion, our study describes the determinants of spatio- Barlow, A.C.D., McDougal, C., Smith, J.L.D., Gurung, B., Bhatta, S.R., Kumal, S., Mahato, temporal variation in abundance and habitat use by Amur tigers along B., Tamang, D.B., 2009. Temporal variation in tiger (Panthera tigris) populations and its implications for monitoring. J. Mammal. 90, 472–478. the China-Russia border. Our results demonstrate that this small, iso- Bhattarai, B.P., Kindlmann, P., 2013. Effect of human disturbance on the prey of tiger in lated tiger population is facing serious hurdles from human activities the Chitwan National Park - implications for park management. J. Environ. Manag. – that are restricting range expansion. To our knowledge, this is the first 131, 343 350. Bischof, R., Broseth, H., Gimenez, O., 2016. Wildlife in a politically divided world: in- comprehensive assessment of the status of Amur tigers in China that sularism inflates estimates of brown bear abundance. Conserv. Lett. 9, 122–130. identifies factors that enhance and limit their distribution. Our findings Broms, K.M., Johnson, D.S., Altwegg, R., Conquest, L.L., 2014. Spatial occupancy models can help policy makers and managers develop a multi-stage plan to applied to atlas data show Southern Ground Hornbills strongly depend on protected areas. Ecol. Appl. 24, 363–374. create a landscape that fosters biodiversity connectivity, local sustain- Brooks, S.P., Gelman, A., 1998. General methods for monitoring convergence of iterative able development, and national and global priorities to increase eco- simulations. J. Comput. Graph. Stat. 7, 434–455. system integrity. Burnham, K.P., Anderson, D.R., 2012. Model Selection and Multimodel Inference. A

277 T. Wang et al. Biological Conservation 217 (2018) 269–279

Practical Information Theoretic Approach. Springer, New York, USA. J.M., Kitchener, A., 2012. Why do Amur tigers maintain exclusive home ranges? Carroll, C., Miquelle, D.G., 2006. Spatial viability analysis of Amur tiger Panthera tigris Relating ungulate seasonal movements to tiger spatial organization in the Russian Far altaica in the Russian Far East: the role of protected areas and landscape matrix in East. J. Zool. 287, 276–282. population persistence. J. Appl. Ecol. 43, 1056–1068. Hughes, J., Haran, M., 2013. Dimension reduction and alleviation of confounding for Carter, N.H., Shrestha, B.K., Karki, J.B., Pradhan, N.M.B., Liu, J.G., 2012. Coexistence spatial generalized linear mixed models. J. R. Stat. Soc. Ser. B (Stat Methodol.) 75, between wildlife and humans at fine spatial scales. Proc. Natl. Acad. Sci. U. S. A. 109, 139–159. 15360–15365. Johnson, D.S., Conn, P.B., Hooten, M.B., Ray, J.C., Pond, B.A., 2013. Spatial occupancy Chanchani, P., Noon, B.R., Bailey, L.L., Warrier, R.A., 2016. Conserving tigers in working models for large data sets. Ecology 94, 801–808. landscapes. Conserv. Biol. 30, 649–660. Jolly, G.M., 1965. Explicit estimates from capture-recapture data with both death and Dave, C., Jhala, Y., 2011. Is competition with livestock detrimental for native wild un- immigration-stochastic model. Biometrika 52, 225–247. gulates? A case study of chital (Axis axis) in Gir Forest, India. J. Trop. Ecol. 27, Joshi, A., Vaidyanathan, S., Mondol, S., Edgaonkar, A., Ramakrishnan, U., 2013. 239–247. Connectivity of tiger (Panthera tigris) populations in the human-influenced forest De Reu, J., Bourgeois, J., Bats, M., Zwertvaegher, A., Gelorini, V., De Smedt, P., Chu, W., mosaic of central India. PLoS One 8, e77980. Antrop, M., De Maeyer, P., Finke, P., Van Meirvenne, M., Verniers, J., Crombe, P., Joshi, A.R., Dinerstein, E., Wikramanayake, E., Anderson, M.L., Olson, D., Jones, B.S., 2013. Application of the topographic position index to heterogeneous landscapes. Seidensticker, J., Lumpkin, S., Hansen, M.C., Sizer, N.C., Davis, C.L., Palminteri, S., Geomorphology 186, 39–49. Hahn, N.R., 2016. Tracking changes and preventing loss in critical tiger habitat. Sci. Dinerstein, E., Loucks, C., Wikramanayake, E., Ginsberg, J., Sanderson, E., Seidensticker, Adv. 2. J., Forrest, J., Bryja, G., Heydlauff, A., Klenzendorf, S., Leimgruber, P., Mills, J., Kanagaraj, R., Wiegand, T., Kramer-Schadt, S., Anwar, M., Goyal, S.P., 2011. Assessing O'Brien, T.G., Shrestha, M., Simons, R., Songer, M., 2007. The fate of wild tigers. habitat suitability for tiger in the fragmented Terai Arc Landscape of India and Nepal. Bioscience 57, 508–514. Ecography 34, 970–981. Dou, H.L., Yang, H.T., Feng, L.M., Mou, P., Wang, T.M., Ge, J.P., 2016. Estimating the Karanth, K.U., Nichols, J.D., 1998. Estimation of tiger densities in India using photo- population size and genetic diversity of Amur tigers in Northeast China. PLoS One 11, graphic captures and recaptures. Ecology 79, 2852–2862. e0154254. Karanth, K.U., Nichols, J.D., Kumar, N.S., Link, W.A., Hines, J.E., 2004. Tigers and their Duangchantrasiri, S., Umponjan, M., Simcharoen, S., Pattanavibool, A., Chaiwattana, S., prey: predicting carnivore densities from prey abundance. Proc. Natl. Acad. Sci. U. S. Maneerat, S., Kumar, N.S., Jathanna, D., Srivathsa, A., Karanth, K.U., 2016. Dynamics A. 101, 4854–4858. of a low-density tiger population in Southeast Asia in the context of improved law Kellner, K., 2016. jagsUI: a wrapper Around rjags to streamline JAGS analyses. R package enforcement. Conserv. Biol. 30, 639–648. version 1.4.2. https://cran.r-project.org/web/packages/jagsUI. Efford, M.G., 2004. Density estimation in live-trapping studies. Oikos 106, 598–610. Kenney, J., Allendorf, F.W., McDougal, C., Smith, J.L.D., 2014. How much gene flow is Efford, M.G., 2016. SECR: spatially explicit capture–recapture. R package version 2.10.3. needed to avoid inbreeding depression in wild tiger populations? Proc. R. Soc. Lond. https://cran.r-project.org/package=secr. B Biol. Sci. 281, 20133337. Efford, M.G., Dawson, D.K., Borchers, D.L., 2009. Population density estimated from lo- Kerley, L.L., Goodrich, J.M., Miquelle, D.G., Smirnov, E.N., Quigley, H.B., Hornocker, cations of individuals on a passive detector array. Ecology 90, 2676–2682. N.G., 2002. Effects of roads and human disturbance on Amur tigers. Conserv. Biol. 16, Fiske, I.J., Chandler, R.B., 2011. Unmarked: an R package for fitting hierarchical models 97–108. of wildlife occurrence and abundance. J. Stat. Softw. 43, 1–23. Kerley, L.L., Mukhacheva, A.S., Matyukhina, D.S., Salmanova, E., Salkina, G.P., Miquelle, Gardner, B., Reppucci, J., Lucherini, M., Royle, J.A., 2010. Spatially explicit inference for D.G., 2015. A comparison of food habits and prey preference of Amur tiger (Panthera open populations: estimating demographic parameters from camera-trap studies. tigris altaica Timminck, 1884) at three sites in the Russian Far East. Integr. Zool. 10, Ecology 91, 3376–3383. 354–364. Gelfand, A.E., Ghosh, S.K., 1998. Model choice: a minimum posterior predictive loss Li, B., Zhang, E.D., Liu, Z.B., 2009a. Livestock depredation by Amur tigers in Hunchun approach. Biometrika 85, 1–11. Nature Reserve, Jilin, China. Acta Theriol. Sin. 29, 231–238. Gelman, A., Meng, X.L., Stern, H., 1996. Posterior predictive assessment of model fitness Li, Z.W., Wu, J.G., Kou, X.J., Tian, Y., Wang, T.M., Mou, P., Ge, J.P., 2009b. Land use via realized discrepancies. Stat. Sin. 6, 733–760. pattern and its dynamics changes in Amur tiger distribution region. Chin. J. Appl. Geweke, J., 1992. Evaluating the accuracy of sampling-based approaches to the calcu- Ecol. 20, 713–724 (in Chinese). lation of posterior moments. In: Bernardo, J.M., Berger, J.O., Dawid, A.P., Smith, Linkie, M., Chapron, G., Martyr, D.J., Holden, J., Leader-Williams, N., 2006. Assessing the A.F.M. (Eds.), Bayesian Statistics 4. Oxford University Press, Oxford, pp. 169–193. viability of tiger subpopulations in a fragmented landscape. J. Appl. Ecol. 43, Gilbert, M., Soutyrina, S.V., Seryodkin, I.V., Sulikhan, N., Uphyrkina, O.V., Goncharuk, 576–586. M., Matthews, L., Cleaveland, S., Miquelle, D.G., 2015. Canine distemper virus as a Linnell, J.D.C., Trouwborst, A., Boitani, L., Kaczensky, P., Huber, D., Reljic, S., Kusak, J., threat to wild tigers in Russia and across their range. Integr. Zool. 10, 329–343. Majic, A., Skrbinsek, T., Potocnik, H., Hayward, M.W., Milner-Gulland, E.J., Goodrich, J.M., Kerley, L.L., Smirnov, E.N., Miquelle, D.G., McDonald, L., Quigley, H.B., Buuveibaatar, B., Olson, K.A., Badamjav, L., Bischof, R., Zuther, S., Breitenmoser, U., Hornocker, M.G., McDonald, T., 2008. Survival rates and causes of mortality of Amur 2016. Border security fencing and wildlife: the end of the transboundary paradigm in tigers on and near the Sikhote-Alin Biosphere Zapovednik. J. Zool. 276, 323–329. Eurasia? PLoS Biol. 14, e1002483. Gopal, R., Qureshi, Q., Bhardwaj, M., Singh, R.K.J., Jhala, Y.V., 2010. Evaluating the Luo, S.J., Kim, J.H., Johnson, W.E., van der Welt, J., Martenson, J., Yuhki, N., Miquelle, status of the endangered tiger Panthera tigris and its prey in Panna Tiger Reserve, D.G., Uphyrkina, O., Goodrich, J.M., Quigley, H.B., Tilson, R., Brady, G., Martelli, P., Madhya Pradesh, India. Oryx 44, 383–389. Subramaniam, V., McDougal, C., Hean, S., Huang, S.Q., Pan, W.S., Karanth, U.K., Harihar, A., Pandav, B., Goyal, S., 2009. Responses of tiger (Panthera tigris) and their prey Sunquist, M., Smith, J.L.D., O'Brien, S.J., 2004. Phylogeography and genetic ancestry to removal of anthropogenic influences in Rajaji National Park, India. Eur. J. Wildl. of tigers (Panthera tigris). PLoS Biol. 2, 2275–2293. Res. 55, 97–105. Mackenzie, D.I., Royle, J.A., 2005. Designing occupancy studies: general advice and al- Harihar, A., Pandav, B., Goyal, S.P., 2011. Responses of leopard Panthera pardus to the locating survey effort. J. Appl. Ecol. 42, 1105–1114. recovery of a tiger Panthera tigris population. J. Appl. Ecol. 48, 806–814. MacKenzie, D.I., Nichols, J.D., Lachman, G.B., Droege, S., Royle, J.A., Langtimm, C.A., Harihar, A., Pandav, B., MacMillan, D.C., 2014. Identifying realistic recovery targets and 2002. Estimating site occupancy rates when detection probabilities are less than one. conservation actions for tigers in a human-dominated landscape using spatially ex- Ecology 83, 2248–2255. plicit densities of wild prey and their determinants. Divers. Distrib. 20, 567–578. Madhusudan, M.D., 2004. Recovery of wild large herbivores following livestock decline Hebblewhite, M., Zimmermann, F., Li, Z., Miquelle, D.G., Zhang, M., Sun, H., Mörschel, in a tropical Indian wildlife reserve. J. Appl. Ecol. 41, 858–869. F., Wu, Z., Sheng, L., Purekhovsky, A., Chunquan, Z., 2012. Is there a future for Amur McLaughlin, K., 2016. Tiger Land. Science 353, 744–745. tigers in a restored tiger conservation landscape in Northeast China? Anim. Conserv. Miquelle, D.G., Smirnov, E.N., Merrill, T.W., Myslenkov, A.E., Quigley, H.B., Hornocker, 15, 579–592. M.G., Schleyer, B.O., 1999. Hierarchical spatial analysis of Amur tiger relationships Hebblewhite, M., Miquelle, D.G., Robinson, H., Pikunov, D.G., Dunishenko, Y.M., to habitat and prey. In: Siedensticker, J., Christie, S., Jackson, P. (Eds.), Riding the Aramilev, V.V., Nikolaev, I.G., Salkina, G.P., Seryodkin, I.V., Gaponov, V.V., Litvinov, Tiger: Tiger Conservation in Human-dominated Landscapes. Cambridge University M.N., Kostyria, A.V., Fomenko, P.V., Murzin, A.A., 2014. Including biotic interactions Press, Cambridge, UK, pp. 71–99. with ungulate prey and humans improves habitat conservation modeling for en- Miquelle, D.G., Pikunov, D.G., Dunishenko, Y.M., Aramilev, V.V., Nikolaev, I.G., Smirnov, dangered Amur tigers in the Russian Far East. Biol. Conserv. 178, 50–64. E.N., Salkina, G.P., Seryodkin, I.V., Gapanov, V.V., Fomenko, P.V., Litvinov, M.N., Henry, P., Miquelle, D., Sugimoto, T., McCullough, D.R., Caccone, A., Russello, M.A., Kostyria, A.V., Korkisko, V.G., Murzin, A.A., 2006. A Survey of Amur (Siberian) 2009. In situ population structure and ex situ representation of the endangered Amur Tigers in the Russian Far East, 2004–2005: Final Report. tiger. Mol. Ecol. 18, 3173–3184. Miquelle, D.G., Goodrich, J.M., Kerley, L.L., Pikunov, D.G., Dunishenko, Y.M., Aramiliev, Hernandez-Blanco, J.A., Rozhnov, V.V., Lukarevskiy, V.S., Naidenko, S.V., Chistopolova, V.V., Nikolaev, I.G., Smirnov, E.N., Salkina, G.P., Endi, Z., Seryodkin, I.V., Carroll, C., M., Sorokin, P.A., Litvinov, M.N., Kotlyar, A.K., 2013. Spatially explicit capture-re- Gapanov, V.V., Fomenko, P.V., Kostyria, A.V., Murzin, A.A., Quigley, H.B., capture method (SECR, SPACECAP): a new approach to determination of the Amur Hornocker, M.G., 2010a. Science-based conservation of Amur tigers in the Russian tiger (Panthera tigris altaica) population density by means of camera traps. Dokl. Biol. Far East and Northeast China. In: Tilson, R., Nyhus, P.J. (Eds.), Tiger of the World: Sci. 453, 365–368. The Science, Politics, and Conservation of Panthera tigris, 2nd edition. Elsevier Hernandez-Blanco, J.A., Naidenko, S.V., Chistopolova, M.D., Lukarevskiy, V.S., Kostyrya, Limited, Oxford, pp. 403–423. A., Rybin, A., Sorokin, P.A., Litvinov, M.N., Kotlyar, A.K., Miquelle, D.G., Rozhnov, Miquelle, D.G., Goodrich, J.M., Smirnov, E.N., Stephens, P.A., Zaumyslova, O.Y., V.V., 2015. Social structure and space use of Amur tigers (Panthera tigris altaica)in Chapron, G., Kerley, L., Murzin, A.A., Hornocker, M.G., Quigley, H.B., 2010b. The Southern Russian Far East based on GPS telemetry data. Integr. Zool. 10, 365–375. Amur tiger: a case study of living on the edge. In: Macdonald, D.W., Loveridge, A.J. Hiby, L., Lovell, P., Patil, N., Kumar, N.S., Gopalaswamy, A.M., Karnath, K.U., 2009. A (Eds.), Biology and Conservation of Wild Felids. Oxford University Press, Oxford, pp. tiger cannot change its stripes: using a three-dimensional model to match images of 325–339. living tigers and tiger skins. Biol. Lett. 5, 383–386. Miquelle, D.G., Rozhnov, V.V., Ermoshin, V., Murzin, A.A., Nikolaev, I.G., Hernandez- Hojnowski, C.E., Miquelle, D.G., Myslenkov, A.I., Strindberg, S., Smirnov, E.N., Goodrich, Blanco, J.A., Naidenko, S.V., 2015. Identifying ecological corridors for Amur tigers

278 T. Wang et al. Biological Conservation 217 (2018) 269–279

(Panthera tigris altaica) and Amur leopards (Panthera pardus orientalis). Integr. Zool. Amur tigers in Hunchun, China: relevance of prey density and implications for pro- 10, 389–402. tected area management. Biol. Conserv. 169, 117–127. Mishra, C., Van Wieren, S.E., Heitkonig, I.M.A., Prins, H.H.T., 2002. A theoretical analysis Sorokin, P.A., Rozhnov, V.V., Krasnenko, A.U., Lukarevskiy, V.S., Naidenko, S.V., of competitive exclusion in a Trans-Himalayan large-herbivore assemblage. Anim. Hernandez-Blanco, J.A., 2016. Genetic structure of the Amur tiger (Panthera tigris Conserv. 5, 251–258. altaica) population: are tigers in Sikhote-Alin and southwest Primorye truly isolated? O'Brien, T.G., Kinnaird, M.F., Wibisono, H.T., 2003. Crouching tigers, hidden prey: Integr. Zool. 11, 25–32. Sumatran tiger and prey populations in a tropical forest landscape. Anim. Conserv. 6, Soutyrina, S.V., Riley, M.D., Goodrich, J.M., Seryodkin, I.V., Miquelle, D.G., 2013. A 131–139. Population Estimate of Amur Tigers Using Camera Traps. Dalnauka, Vladivostok. Petrunenko, Y.K., Montgomery, R.A., Seryodkin, I.V., Zaumyslova, O.Y., Miquelle, D.G., Tian, Y., Wu, J.G., Kou, X.J., Li, Z.W., Wang, T.M., Mou, P., Ge, J.P., 2009. Macdonald, D.W., 2016. Spatial variation in the density and vulnerability of pre- Spatiotemporal pattern and major causes of the Amur tiger population dynamics. ferred prey in the landscape shape patterns of Amur tiger habitat use. Oikos 125, Biodivers. Sci. 17, 211–225 (in Chinese). 66–75. Tian, Y., Wu, J.G., Smith, A.T., Wang, T.M., Kou, X.J., Ge, J.P., 2011. Population viability Pettigrew, M., Xie, Y., Kang, A.L., Rao, M., Goodrich, J., Liu, T., Berger, J., 2012. Human- of the Siberian tiger in a changing landscape: going, going and gone? Ecol. Model. carnivore conﬂict in China: a review of current approaches with recommendations 222, 3166–3180. for improved management. Integr. Zool. 7, 210–226. Walston, J., Robinson, J.G., Bennett, E.L., Breitenmoser, U., da Fonseca, G.A.B., Goodrich, Pettorelli, N., Owen, H.J.F., Duncan, C., 2016. How do we want Satellite Remote Sensing J., Gumal, M., Hunter, L., Johnson, A., Karanth, K.U., Leader-Williams, N., to support biodiversity conservation globally? Methods Ecol. Evol. 7, 656–665. MacKinnon, K., Miquelle, D., Pattanavibool, A., Poole, C., Rabinowitz, A., Smith, Pitman, R.T., Fattebert, J., Williams, S.T., Williams, K.S., Hill, R.A., Hunter, L.T.B., J.L.D., Stokes, E.J., Stuart, S.N., Vongkhamheng, C., Wibisono, H., 2010. Bringing the Robinson, H., Power, J., Swanepoel, L., Slotow, R., 2017. Cats, connectivity and tiger back from the brink-the six percent solution. PLoS Biol. 8, e1000485. conservation: incorporating data sets and integrating scales for wildlife management. Wang, H.R., Wang, T.M., Ge, J.P., 2012. Variation trends of the vegetations in distribution J. Appl. Ecol. http://dx.doi.org/10.1111/1365-2664.12851. region of Amur tiger based on MODIS NDVI. Chin. J. Appl. Ecol. 23, 2821–2828. Rayan, D.M., Linkie, M., 2015. Conserving tigers in Malaysia: a science-driven approach Wang, T.M., Yang, H.T., Xiao, W.H., Feng, L.M., Mou, P., Ge, J.P., 2014. Camera traps for eliciting conservation policy change. Biol. Conserv. 184, 18–26. reveal Amur tiger breeding in NE China. Cat News 61, 18–19. Ridout, M.S., Linkie, M., 2009. Estimating overlap of daily activity patterns from camera Wang, F., McShea, W.J., Wang, D.J., Li, S., 2015a. Shared resources between giant panda trap data. J. Agric. Biol. Environ. Stat. 14, 322–337. and sympatric wild and domestic mammals. Biol. Conserv. 186, 319–325. Royle, J.A., Dorazio, R.M., 2012. Parameter-expanded data augmentation for Bayesian Wang, T.M., Feng, L.M., Mou, P., Ge, J.P., Li, C., Smith, J.L.D., 2015b. Long-distance analysis of capture-recapture models. J. Ornithol. 152, S521–S537. dispersal of an Amur tiger indicates potential to restore the North-east China/Russian Royle, J.A., Karanth, K.U., Gopalaswamy, A.M., Kumar, N.S., 2009. Bayesian inference in Tiger Landscape. Oryx 49, 578–579. camera trapping studies for a class of spatial capture-recapture models. Ecology 90, Wang, T.M., Feng, L.M., Mou, P., Wu, J.G., Smith, J.L.D., Xiao, W.H., Yang, H.T., Dou, 3233–3244. H.L., Zhao, X.D., Cheng, Y.C., Zhou, B., Wu, H.Y., Zhang, L., Tian, Y., Guo, Q.X., Kou, Royle, J.A., Fuller, A.K., Sutherland, C., 2016. Spatial capture-recapture models allowing X.J., Han, X.M., Miguelle, D.G., Oliver, C.D., Xu, R.M., Ge, J.P., 2016. Amur tigers Markovian transience or dispersal. Popul. Ecol. 58, 53–62. and leopards returning to China: direct evidence and a landscape conservation plan. Sappington, J.M., Longshore, K.M., Thompson, D.B., 2007. Quantifying landscape rug- Landsc. Ecol. 31, 491–503. gedness for animal habitat analysis: a case study using bighorn sheep in the Mojave Wang, T.M., Feng, L.M., Yang, H.T., Han, B.Y., Zhao, Y.H., Juan, L., Lu, X.Y., Zou, L., Li, Desert. J. Wildl. Manag. 71, 1419–1426. T., Xiao, W.H., Mou, P., Smith, J.L.D., Ge, J.P., 2017. A science-based approach to Schuster, R., Arcese, P., 2013. Using bird species community occurrence to prioritize guide Amur leopard recovery in China. Biol. Conserv. 210, 47–55. forests for old growth restoration. Ecography 36, 499–507. Xiao, W.H., Feng, L.M., Zhao, X.D., Yang, H.T., Dou, H.L., Cheng, Y.C., Mou, P., Wang, Seber, G., 1965. A note on the multiple-recapture census. Biometrika 52, 249–259. T.M., Ge, J.P., 2014. Distribution and abundance of Amur tiger, Amur leopard and Smith, J.L.D., McDougal, C.W., Sunquist, M.E., 1987. Female land tenure system in tigers. their ungulate preys in Hunchun National Nature Reserve, Jilin. Biodivers. Sci. 22, In: Tilson, R.L., Seal, U.S. (Eds.), Tigers of the World: The Biology, Biopolitics, 717–724 (in Chinese). Management and Conservation of an Endangered Species. Noyes Publications, Park Xiao, W.H., Feng, L.M., Mou, P., Miquelle, D.G., Hebblewhite, M., Robinson, H.S., Zhao, Ridge, New Jersey, pp. 97–108. X.D., Zhou, B., Wang, T.M., Ge, J.P., 2016. Estimating abundance and density of Smith, J.L.D., Ahearn, S.C., McDougal, C., 1998. Landscape analysis of tiger distribution Amur tigers along the Sino-Russian border. Integr. Zool. 11, 322–332. and habitat quality in Nepal. Conserv. Biol. 12, 1338–1346. Yumnam, B., Jhala, Y.V., Qureshi, Q., Maldonado, J.E., Gopal, R., Saini, S., Srinivas, Y., Soh, Y.H., Carrasco, L.R., Miquelle, D.G., Jiang, J.S., Yang, J., Stokes, E.J., Tang, J.R., Fleischer, R.C., 2014. Prioritizing tiger conservation through landscape genetics and Kang, A.L., Liu, P.Q., Rao, M., 2014. Spatial correlates of livestock depredation by habitat linkages. PLoS One 9, e111207.

279

View publication stats